A brief yacc tutorial

1
A brief yacc tutorial

• Saumya Debray
• The University of Arizona
• Tucson, AZ 85721

2
Yacc Overview

• Parser generator
• Takes a specification for a context-free grammar.
• Produces code for a parser.

Output C code implementing a parser function
yyparse() file default y.tab.c
Input a set of grammar rules and actions
yacc (or bison)
3
Scanner-Parser interaction

• Parser assumes the existence of a function int
  yylex() that implements the scanner.
• Scanner
• return value indicates the type of token found
• other values communicated to the parser using
  yytext, yylval (see man pages).
• Yacc determines integer representations for
  tokens
• Communicated to scanner in file y.tab.h
• use yacc -d to produce y.tab.h
• Token encodings
• end of file represented by 0
• a character literal its ASCII value
• other tokens assigned numbers ? 257.

4
Using Yacc
lexical rules
grammar rules
y.output
flex
yacc
describes states, transitions of parser
(useful for debugging)
yacc -v
yacc -d ? y.tab.h
lex.yy.c
y.tab.c
yylex()
yyparse()
tokens
parsed input
input
5
int yyparse()

• Called once from main() user-supplied
• Repeatedly calls yylex() until done
• On syntax error, calls yyerror() user-supplied
• Returns 0 if all of the input was processed
• Returns 1 if aborting due to syntax error.
• Example
• int main() return yyparse()

6
yacc input format

• A yacc input file has the following structure

• definitions
• rules
• user code

required
optional
Shortest possible legal yacc input

7
Definitions

• Information about tokens
• token names
• declared using token
• single-character tokens dont have to be declared
• any name not declared as a token is assumed to be
  a nonterminal.
• start symbol of grammar, using start
  optional
• operator info
• precedence, associativity
• stuff to be copied verbatim into the output
  (e.g., declarations, includes) enclosed in
  

8
Rules

• Rule RHS can have arbitrary C code embedded,
  within . E.g.
• A B1 printf(after B1\n) x 0 B2 x
  B3
• Left-recursion more efficient than
  right-recursion
• A A x rather than A x A

9
Conflicts

• Conflicts arise when there is more than one way
  to proceed with parsing.
• Two types
• shift-reduce default action shift
• reduce-reduce default reduce with the first
  rule listed
• Removing conflicts
• specify operator precedence, associativity
• restructure the grammar
• use y.output to identify reasons for the conflict.

10
Specifying Operator Properties

• Binary operators left, right, nonassoc
• left '' '-'
• left '' '/'
• right '
• Unary operators prec
• Changes the precedence of a rule to be that of
  the token specified. E.g.
• left '' '-'
• left '' '/
• Expr expr expr
• expr prec

Operators in the same group have the same
precedence
11
Specifying Operator Properties

• Binary operators left, right, nonassoc
• left '' '-'
• left '' '/'
• right ''
• Unary operators prec
• Changes the precedence of a rule to be that of
  the token specified. E.g.
• left '' '-'
• left '' '/'
• Expr expr expr
• expr prec

Operators in the same group have the same
precedence
Across groups, precedence increases going down.
12
Specifying Operator Properties

• Binary operators left, right, nonassoc
• left '' '-'
• left '' '/'
• right '
• Unary operators prec
• Changes the precedence of a rule to be that of
  the token specified. E.g.
• left '' '-'
• left '' '/'
• Expr expr '' expr
• '' expr prec ''

Operators in the same group have the same
precedence
Across groups, precedence increases going down.
The rule for unary has the same (high)
precedence as
13
Error Handling

• The token error is reserved for error
  handling
• can be used in rules
• suggests places where errors might be detected
  and recovery can occur.
• Example
• stmt IF '(' expr ')' stmt
• IF '(' error ')' stmt
• FOR

Intended to recover from errors in expr
14
Parser Behavior on Errors

• When an error occurs, the parser
• pops its stack until it enters a state where the
  token error is legal
• then behaves as if it saw the token error
• performs the action encountered
• resets the lookahead token to the token that
  caused the error.
• If no error rules specified, processing halts.

15
Controlling Error Behavior

• Parser remains in error state until three tokens
  successfully read in and shifted
• prevents cascaded error messages
• if an error is detected while parser in error
  state
• no error message given
• input token causing the error is deleted.
• To force the parser to believe that an error has
  been fully recovered from
• yyerrok
• To clear the token that caused the error
• yyclearin

16
Placing error tokens

• Some guidelines
• Close to the start symbol of the grammar
• To allow recovery without discarding all input.
• Near terminal symbols
• To allow only a small amount of input to be
  discarded on an error.
• Consider tokens like ), , that follow
  nonterminals.
• Without introducing conflicts.

17
Error Messages

• On finding an error, the parser calls a function
• void yyerror(char s) / s points to an error
  msg /
• user-supplied, prints out error message.
• More informative error messages
• int yychar token no. of token causing the error.
• user program keeps track of line numbers, as well
  as any additional info desired.

18
Error Messages example

• include "y.tab.h"
• extern int yychar, curr_line
• static void print_tok()
• if (yychar lt 255)
• fprintf(stderr, "c", yychar)
• else
• switch (yychar)
• case ID
• case INTCON

• void yyerror(char s)
• fprintf(stderr,
• "line d s",
• curr_line,
• s)
• print_tok()

19
Debugging the Parser

• To trace the shift/reduce actions of the parser
• when compiling
• define YYDEBUG
• at runtime
• set yydebug 1 / extern int yydebug /

20
Adding Semantic Actions

• Semantic actions for a rule are placed in its
  body
• an action consists of C code enclosed in
• may be placed anywhere in rule RHS
• Example
• expr ID symTbl_lookup(idname)
• decl type_name tval id_list

21
Synthesized Attributes

• Each nonterminal can return a value
• The return value for a nonterminal X is
  returned to a rule that has X in its body,
  e.g.
• A X
• X
• This is different from the values returned by the
  scanner to the parser!

value returned by X
22
Attribute return values

• To access the value returned by the ith symbol in
  a rule body, use i
• an action occurring in a rule body counts as a
  symbol. E.g.
• decl type tval 1 id_list
  symtbl_install(3, tval)
• To set the value to be returned by a rule, assign
  to
• by default, the value of a rule is the value of
  its first symbol, i.e., 1.

1
2
3
4
23
Example

• / A variable declaration is an identifier
  followed by an optional subscript, e.g., x or
  x10 /

• var_decl ident opt_subscript
• if ( symtbl_lookup(1) ! NULL )
• ErrMsg(multiple
  declarations, 1)
• else
• st_entry symtbl_install(1)
• if ( 2 ARRAY )
• st_entry-gtbase_type
  ARRAY
• st_entry-gtelement_type
  tval
• else
• st_entry-gtbase_type tval
• st_entry-gtelement_type
  UNDEF

• opt_subscript INTCON ARRAY
• / null /
  INTEGER

24
Declaring Return Value Types

• Default type for nonterminal return values is
  int.
• Need to declare return value types if nonterminal
  return values can be of other types
• Declare the union of the different types that may
  be returned
• union
• symtbl_ptr st_ptr
• id_list_ptr list_of_ids
• tree_node_ptr syntax_tree_ptr
• int value
• Specify which union member a particular grammar
  symbol will return
• token ltvaluegt INTCON, CHARCON
• type ltst_ptrgt identifier
• type ltsyntax_tree_ptrgt expr, stmt

terminals
nonterminals
25
Conflicts

• A conflict occurs when the parser has multiple
  possible actions in some state for a given next
  token.
• Two kinds of conflicts
• shift-reduce conflict
• The parser can either keep reading more of the
  input (shift action), or it can mimic a
  derivation step using the input it has read
  already (reduce action).
• reduce-reduce conflict
• There is more than one production that can be
  used for mimicking a derivation step at that
  point.

26
Example of a conflict

• Grammar rules
• S ? if ( e ) S / 1 /
  Input if ( e1 ) if ( e2 ) S2 else S3
• if ( e ) S else S / 2 /
• Parser state when input token else
• Input already seen if ( e1 ) if ( e2 ) S2
• Choices for continuing

• 1. keep reading input (shift)
• else part of innermost if
• eventual parse structure
• if (e1) if (e2) S2 else S3

• 2. mimic derivation step using
• S ? if ( e ) S (reduce)
• else part of outermost if
• eventual parse structure
• if (e1) if (e2) S2 else S3

shift-reduce conflict
27
Handling Conflicts

• General approach
• Iterate as necessary
• Use yacc -v to generate the file y.output.
• Examine y.output to find parser states with
  conflicts.
• For each such state, examine the items to figure
  why the conflict is occurring.
• Transform the grammar to eliminate the conflict

28
Understanding y.output
S 0 S 0 / epsilon /
Input grammar
y.output file
0 accept S end 1 S 0 S 0 2
1 shift/reduce conflict (shift 1, reduce 2)
on 0 state 1 .
artificial rule, introduced by yacc
rules from input grammar
rule numbers
indicates which state has conflict
information about state 1
29
Understanding y.output contd
no. of state with conflict

• 1 shift/reduce conflict (shift 1, reduce 2) on
  0

• nature of the conflict
• shift 1 shift and go to state 1
• reduce 2 reduce using rule 2 (S ? ?)

input token on which the conflict occurs
Whats going on?

• havent reached midpoint of input string
• seen 0, looking to process S
• ? if input has 0, continue processing
• ? shift

state 1 S 0 . S 0 S .

• Reached midpoint of input string
• next 0 in input is from second half of string
• if input is 0, reduce using (S ? ?)